The AUTO DETECT and INTERACTIVE DETECT modes have the potential to be real time savers. Adding all those correspondence points by hand takes a long time. My suggestion however is to try it on more models.
Example of “auto detect”:
Example of “interactive detect”:
This is what it should have been:
Note: I could not move point 5 for some reason until I zoomed in and out, panned around tried lots of things, moved other points, etc. Suddenly it started moving. This sort of unpredictable behavior is annoying.
Also, when I turn to “wireframe” for some reason transfer_character/face_skin is not visible. I turn on the original face_skin layer again to see the mesh.
Why am I using the wireframe mode? Sometimes it goes into a mode of “I won’t recognize all mesh vertices”, but I was trying to prove it. With wireframe I can take a screenshot proving sometimes it rejects some vertices.
(I then went on a tried clicking interactive detect the second time, and it seemed to do a decent job after I manually corrected the first set of points. I tried then to view blend shapes using those points, and a2f crashed.)